AI agents

Give your agent the reference for the operation it is implementing, from one of three sources:

The docs MCP server searches the docs and returns endpoint schemas as the agent needs them.
/SKILL.md holds the public HTTP API reference in one file for offline lookup.
Any docs page as Markdown. Append .md to its URL. The OpenAPI spec covers structured schema access.

These are read-only. To let an agent create and inspect resources in your account, use the Flint CLI.

Docs MCP server#

Point an MCP-capable agent at the Flint docs and it can search them, read any page, and pull the full request and response schema for any endpoint, so it works from the current API surface instead of guessing from memory:

Text

https://developers.withflintpay.com/mcp

The server covers documentation only. It does not touch your Flint account, your data, or the payments API, so no Flint API key is involved.

The server exposes these read-only tools:

Tool	What it does
`search_docs`	Search guides, API reference, and blog. Returns matching pages with URLs.
`search_flint_help`	Search public Flint Help threads for answered questions, known issues, and feedback.
`get_page`	Fetch one docs page as Markdown by its path (for example `/docs/guides/webhooks`).
`list_endpoints`	List public API endpoints (method, path, summary), optionally filtered by resource (for example `orders`).
`get_endpoint`	Get the full reference for one endpoint by its `operationId`, including request and response schemas and code samples.

A typical agent loop is search_docs or list_endpoints to find the right operation, then get_endpoint to read its exact schema before writing the call.

Connect an agent#

Add the server with the Claude Code CLI:

Shell

claude mcp add --transport http flint-docs https://developers.withflintpay.com/mcp

Then ask Claude to work against the Flint API. It will call the Flint docs tools when it needs endpoint details.

Add the server to your MCP config (.cursor/mcp.json):

JSON

{
  "mcpServers": {
    "flint-docs": {
      "url": "https://developers.withflintpay.com/mcp"
    }
  }
}

The transport is stateless JSON-RPC 2.0 over HTTP POST. List the tools directly:

cURL

curl -s https://developers.withflintpay.com/mcp \
  -H "Content-Type: application/json" \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/list"}'

Call one:

cURL

curl -s https://developers.withflintpay.com/mcp \
  -H "Content-Type: application/json" \
  -d '{"jsonrpc":"2.0","id":2,"method":"tools/call","params":{"name":"search_docs","arguments":{"query":"idempotency"}}}'

Protocol notes#

Transport. Stateless Streamable HTTP: JSON-RPC 2.0 over POST. There is no session or server-initiated stream, so a GET returns 405.
Protocol version. The server advertises 2025-03-26 and also accepts 2024-11-05 and 2025-06-18 clients.
Batching. Batched requests are supported, capped at 20 messages per call.
Auth. None. The endpoint is public and read-only.

The MCP server reads the same docs corpus as this site, so it always reflects the current API surface. The full corpus is also available as a single file at /llms-full.txt.

SKILL.md#

The file is a self-contained markdown document covering:

Auth & base URL: the public API host plus test/live auth conventions
Public HTTP API surface: /v1 endpoints and request/response guidance
Conventions: money fields, ID prefixes, pagination, sparse updates and patch objects, metadata limits
State machines: order, payment, invoice, and subscription lifecycles
Errors: structured error format and retry behavior
Quick start examples: 5 curl commands covering the most common flows

Use the endpoint schema for exact fields and constraints. Keep the full reference available for offline lookup without adding it to every task's starting context.

The SKILL.md file tracks the same API surface as these docs. When new endpoints or fields ship, both are updated together.

Download it for offline work:

Text

https://developers.withflintpay.com/SKILL.md

Use SKILL.md for a fixed reference and the MCP server for interactive lookups. They are not mutually exclusive.

Point your agent at Flint#

Add a scoped reference to your project's CLAUDE.md:

Markdown

# Flint Pay API

For Flint API integration changes, look up the affected endpoint and relevant guide
through https://developers.withflintpay.com/mcp, or fetch the page's Markdown by
appending .md to its URL. Use https://developers.withflintpay.com/SKILL.md for
offline reference, reading only the relevant conventions and resource sections.

Keep the reference tied to API integration work so unrelated edits need no Flint docs.

Add to your project's AGENTS.md for OpenAI Codex:

Markdown

# Flint Pay API

For Flint API integration changes, look up the affected endpoint and relevant guide
through https://developers.withflintpay.com/mcp, or fetch the page's Markdown by
appending .md to its URL. Use https://developers.withflintpay.com/SKILL.md for
offline reference, reading only the relevant conventions and resource sections.

This gives Codex a lookup path when a change needs Flint API details.

Use the same scoped instruction in your project's agent rules:

Markdown

# Flint Pay API

For Flint API integration changes, retrieve the affected endpoint and guide from
https://developers.withflintpay.com/mcp or the relevant docs page's .md URL.

Fetch the offline reference when your workflow needs it:

cURL

curl -fsS https://developers.withflintpay.com/SKILL.md -o flint-api-reference.md

Extract the relevant conventions and resource sections for the task, or let the agent search the saved file. Load the full text only when the requested coverage needs it.

Let an agent call the API#

SKILL.md and the docs MCP server give an agent knowledge. To let it act, the Flint CLI is the shortest path: it exposes the whole command surface as machine-readable JSON Schema, so an agent can discover a command, build a valid request body, and validate it before spending a call.

Shell

flint schema commands --output json
flint schema input payment-intents.create --output json

--output json also forces non-interactive mode, so no command blocks on a prompt. Agents that speak MCP can skip the shell entirely with flint mcp serve, which serves the CLI's commands as MCP tools over stdio. Sign-in, key import, and context selection stay in the terminal; see Agents and MCP.

When an agent gets stuck, flint help search looks for an existing answer on Flint Help without a credential, and flint support open --ai-agent --no-open returns a composer link a person can review and post. --ai-agent labels the thread so whoever answers knows an agent wrote it. The CLI ships no command that submits a report on its own, so nothing leaves the machine unattended. See Getting help.